Preformulation Studies and Characterization of the Physicochemical Properties of Amorphous Polymers Using Artificial Neural Networks

The utility of artificial neural networks (ANNs) as a preformulation tool to determine the physicochemical properties of amorphous polymers such as the hydration characteristics, glass transition temperatures and rheological properties was investigated. The neural network simulator, CAD/Chem, based on the delta back-propagation paradigm was used for this study. The ANNs software was trained with sets of experimental data consisting of different polymer blends with known water-uptake profiles, glass transition temperatures and viscosity values. A set of similar data, not initially exposed to the ANNs was used to validate the ability of the ANNs to recognize patterns. The results of this investigation indicate that the ANNs accurately predicted the water-uptake, glass transition temperatures and viscosities of different amorphous polymers and their physical blends with a low % error (0-8%) of prediction. The ANNs also showed good correlation between the water-uptake and changes in the glass transition temperatures of the polymers. This study demonstrated the potential of the ANNs as a preformulation tool to evaluate the characteristics of amorphous polymers. This is particularly relevant when designing sustained release formulations that require the use of a fast hydrating polymer matrix.